580049 - allow to disable or control DOM GC/CC heuristics for testing purposes

Status: NEW

(Core :: DOM: Core & HTML, defect, P5)

Comment 1

[Andreas Gal :gal]

Currently the JS callback drives the CC heuristics. It should be possible to manipulate and in particular disable the heuristics for testing purposes.

Comment 2

[Olli Pettay [:smaug][[e-mail]]]

What heuristics you're talking about?
User activity observer drives CC currently.

Adding some pref for testing should easy.

Comment 3

[Andreas Gal :gal]

This might have changed recently then. Either way, yes, exactly thats what we need to test this. Disable all the heuristics while we build up some garbage and then GC and CC as needed (I am about to land a patch that allows to explicitly trigger GC and CC separately).

Comment 4

[alice nodelman [:alice] [:anode]]

Any movement here?  This is blocking full development of the test described in in bug 572542.

Comment 5

[Andreas Gal :gal]

bent is working on this

Comment 6

[alice nodelman [:alice] [:anode]]

Any update?

Comment 7

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Oh, yikes. I've been off in the weeds, not sure when I can get back to this. Is this blocking you?

Comment 8

[alice nodelman [:alice] [:anode]]

Yeah, the creation of the test is blocked on having the ability to switch on/off automatic memory management so that we can create a controlled test environment.

Comment 9

[Olli Pettay [:smaug][[e-mail]]]

So what all needs the on/off switch? CC only? Also GC? Or CC and GC when triggered
by DOM code?

Comment 10

[alice nodelman [:alice] [:anode]]

We would like to be able to disable CC + GC, load a bunch of things in browser and then force a CC or GC and observe how long it takes.  So, mostly disable the default behavior and be able to create CC/GC requests in controlled situations so that they can be timed.

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

I don't know if it is possible to really disable gc.
Adding a pref to disable cc would be easy.

Comment 12

[alice nodelman [:alice] [:anode]]

If GC occurs in a non-controlled fashion I doubt that we'll be able to get any reasonable numbers out of any CC/GC tests - since you'll never know if you just did a cleanup before requesting one.

Comment 13

[Firefox Bug Husbandry Bot]

https://bugzilla.mozilla.org/show_bug.cgi?id=1472046

Move all DOM bugs that haven’t been updated in more than 3 years and has no one currently assigned to P5.

If you have questions, please contact :mdaly.